Training bug

[gyupro]

Hi, thanks for you wonderful work.

I trained vision-encoder-decoder model using your framework manga_ocr_dev, and saved the model using the script.

and load the model failed as it's shown above, it's not an error but warnings.

and my output structure is different from yours, I don't have tokenizer generated, but you have. I was wondering what I have done wrong. I am using transformers 4.43.3

my output

your output